A High-Resolution Linkage Map Construction and QTL Analysis for Morphological Traits in Anthurium (Anthurium andraeanum Linden)

Anthurium andraeanum Linden is a prominent ornamental plant belonging to the family Araceae and is cultivated worldwide. The morphology characteristics are crucial because they significantly impact ornamental values, commercial properties, and the efficiency of space utilization in production. However, only a few related investigations have been conducted in anthurium to date. In this study, an F1 genetic segregation population containing 160 progenies was generated through hybridization between potted and cut anthurium varieties. Fifteen morphological traits were assessed and revealed substantial levels of genetic variation and widespread positive correlation. Based on specific length amplified fragment (SLAF) sequencing technology, 8171 single nucleotide polymorphism (SNP) markers were developed, and the high-density linkage map of 2202.27 cM in length distributed on 15 linkage groups was constructed successfully, with an average distance of 0.30 cM. Using the inclusive composite interval mapping (ICIM) method, 59 QTLs related to 15 key morphological traits were successfully identified, which explained phenotypic variance (PVE) ranging from 6.21% to 17.74%. Thirty-three of those associated with 13 traits were designated as major QTLs with PVE > 10%. These findings offer valuable insights into the genetic basis of quantitative traits and are beneficial for molecular marker-assisted selection (MAS) in anthurium breeding.


Introduction
Anthurium Schott is the largest Neotropical genus of the family Araceae, composed of more than 950 genera and 2000 species, with a wide range of morphological diversity [1].Anthurium is a flowering plant species, famous for its exotic shape, colorful spathe, and continuous blooming, that has been grown as a cut or potted flower worldwide [2,3].To date, genetic research has been conducted mainly focused on the spathe color, blight resistance, and postharvest vase life [4][5][6][7].Anthurium also shows a wide genetic diversity in morphological traits related to spathe, spadix, and leaf, which not only have a significant impact on its ornamental value but also determine the effective utilization of greenhouse space in production [8,9].So, it is promising to develop novel varieties with distinct appearance structures to address market demands.However, due to the limited understanding of the genetic mechanisms underlying the complex morphology, breeding for morphological traits of anthurium has been time-consuming and costly.
In general, morphological traits are controlled by multiple genes, which can be identified through quantitative trait locus (QTL) mapping [10,11].A series of genetic linkage maps were constructed and utilized for mapping various QTLs in ornamental plants, such as rose [12,13], carnation [14], chrysanthemum [15], lily [16,17], petunia [18], and tree pony [19].Numerous linkage maps were constructed via traditional molecule markers such as RAPDs, SSRs, and AFLPs, which were limited by the number of reliable, repeatable, and stable markers [20][21][22][23].In recent years, with the development of next-generation sequencing (NGS), simplified genome technologies, including the specific length amplified fragment sequencing (SLAF-seq), have been widely used for the development of single nucleotide polymorphism (SNP) markers and the construction of linkage maps [24].To date, genetic linkage maps of many ornamental plants have been constructed using SLAFseq technology, even polypoid plants with high heterozygosity and without reference genomes [25][26][27][28].
However, it is also challenging to construct a linkage map for anthurium due to the long generations, absence of pure lines, and complex heterozygosity generated by interspecific hybridization [29].There is now only one genetic map available, which comprises 228 markers, including 99 RAPDs, 21 ISSRs, and 108 SRAPs, based on the interspecific F 1 populations of 43 individuals descended from A. ornatum Schott and A. andraeanum Linden.The linkage map of A. ornatum was 1233.5 cM in length, distributed in 10 linkage groups (LGs), whereas the linkage map of A. andraeanum was 1023.5 cM in length, distributed in 12 LGs.The LGs were inconsistent with the haploid chromosome number of the Anthurium species (n = 15), and they only covered 77% and 73% of the genomes of A. ornatum and A. andraeanum, respectively [30].Therefore, it is imperative to develop a genetic linkage map with a higher resolution to facilitate genetics and genomics research in anthurium.
In this study, SNP markers based on SLAF-seq technology were developed and the first high-density genetic linkage map of anthurium was constructed.Based on the map, QTLs related to 15 morphological traits were identified.It provides helpful genetics tools and information for molecular marker-assisted selection (MAS) in anthurium breeding.

Phenotypic Analysis
Parameters of 15 morphological traits for the parents and F 1 progenies were measured and statistically analyzed (Figure 1).The coefficients of variation (CV) of traits showed a high degree of genetic variation in the progenies (Table 1).The spathe left distance (LED) and right ear distance (RED), with the CV values of 34.04 and 35.53%, were significantly higher than other traits.Followed by spadix length (SpdL), spathe length (SptL), pedicel diameter (PedD), petiole length (PetL), spathe width (SpdW), and pedicel length (PedL), with values of 26.13%, 24.27%, 23.88%, 22.22%, 22.11%, and 21.82%, respectively.The CV of plant height (PH), leaf length (LL), leaf width (LW), and spadix diameter (top (SpdTD), middle (SpdMD), and base (SpdBD)) were all below 20%.Correlation analysis was conducted among 15 traits (Figure 2).The results indicated that the length, comprising spadix-, pedicel-, and petiole-, were significantly positively correlated with their diameter.The spathe length was strongly positively correlated with its width (r 2 = 0.92, p < 0.001), the distance of the left and right ear in spathe was strongly positively correlated (r 2 = 0.82, p < 0.001), and the leaf length was strongly positively correlated with its width (r 2 = 0.89, p < 0.001).While the plant height was positively correlated with the length of pedicel and petiole.The absolute values of skewness and kurtosis of 15 morphological traits in the 160 progenies were less than one with normal distributions.They were typical quantitative traits, which were suitable for QTL analysis (Figure 3).The absolute values of skewness and kurtosis of 15 morphological traits in the 160 progenies were less than one with normal distributions.They were typical quantitative traits, which were suitable for QTL analysis (Figure 3).The absolute values of skewness and kurtosis of 15 morphological traits in the 160 progenies were less than one with normal distributions.They were typical quantitative traits, which were suitable for QTL analysis (Figure 3).

SLAF Sequencing Data Analysis and Genotyping
To genotype 'Pink Champion' (♀), 'Acropolis' (♂), and F 1 progenies, SLAF-seq was performed and 188.37 Gb of raw data were obtained with a Q30 of 95.81% and GC content of 40.5%.The number of reads for the female and male parents was 9,869,344 and 11,383,208, and the mean for the F 1 progeny was 5,790,723.The average sequencing depth was 80.65× for 'Pink Champion', 91.85× for 'Acropolis', and 31.68× for the F 1 progeny (Table 2).Among these reads, 327,963 SNP markers were identified in all, of which 131,951 were successfully encoded and genotyped into eight segregation patterns (ab × cd, ef × eg, lm × ll, nn × np, aa × bb, hk × hk, cc × ab, ab × cc) (Figure 4).Filtered out 31,971 SNP markers were classified into the pattern of aa × bb (in the ratio of 24.23%), which is inapplicable to the CP model; the remaining 99,980 markers could be used for genetic map construction.To ensure a high-quality genetic map, low-quality SNP markers with integrity lower than 85%, parental information missing, and segregation separation p < 0.05 were removed.Finally, 10,648 SNP markers were identified for the genetic map construction.

SLAF Sequencing Data Analysis and Genotyping
To genotype 'Pink Champion' (♀), 'Acropolis' (♂), and F1 progenies, SLAF-seq was performed and 188.37 Gb of raw data were obtained with a Q30 of 95.81% and GC content of 40.5%.The number of reads for the female and male parents was 9,869,344 and 11,383,208, and the mean for the F1 progeny was 5,790,723.The average sequencing depth was 80.65× for 'Pink Champion', 91.85× for 'Acropolis', and 31.68× for the F1 progeny (Table 2).Among these reads, 327,963 SNP markers were identified in all, of which 131,951 were successfully encoded and genotyped into eight segregation patterns (ab × cd, ef × eg, lm × ll, nn × np, aa × bb, hk × hk, cc × ab, ab × cc) (Figure 4).Filtered out 31,971 SNP markers were classified into the pattern of aa × bb (in the ratio of 24.23%), which is inapplicable to the CP model; the remaining 99,980 markers could be used for genetic map construction.To ensure a high-quality genetic map, low-quality SNP markers with integrity lower than 85%, parental information missing, and segregation separation p < 0.05 were removed.Finally, 10,648 SNP markers were identified for the genetic map construction.

High-Density Genetic Map Construction
The modified logarithm of odds (MLOD) values were calculated between two SNP markers, and fewer than 10 were removed.A total of 8171 SNP markers (in the ratio of 76.48%) were ultimately retained for the high-density genetic map construction, which were distributed into 15 linkage groups (LGs).The total map distance of female and male parents was 2176.58 cM and 1940.36 cM, respectively.By integrating the parents' genetic map, a linkage map with 2202.27cM in length with an average distance of 0.30 cM was constructed (Figure 5 and Table 3).The largest linkage group was LG6 with a length of

High-Density Genetic Map Construction
The modified logarithm of odds (MLOD) values were calculated between two SNP markers, and fewer than 10 were removed.A total of 8171 SNP markers (in the ratio of 76.48%) were ultimately retained for the high-density genetic map construction, which were distributed into 15 linkage groups (LGs).The total map distance of female and male parents was 2176.58 cM and 1940.36 cM, respectively.By integrating the parents' genetic map, a linkage map with 2202.27cM in length with an average distance of 0.30 cM was constructed (Figure 5 and Table 3).The largest linkage group was LG6 with a length of 226.98 cM which harbored 1082 markers, while the smallest linkage group was LG3 with 68.23  LG12, which harbored 987 markers covering a length of 85.27 cM with the least average interval of 0.09 cM.The max gap on LG12 was only 4.46 cM, which is smaller than that of other linkage groups, while the largest genetic gap was found in LG14, with 23.88 cM genetic distance.The percentage of gap (<5 cM) was 98.40%, indicating that the markers were relatively well-distributed in the map [31].According to a chi-square test (p < 0.05) of the 8171 SNP markers, 429 of which (with a ratio of 5.25%) were segregation distortion markers.The greatest number of segregation distortion markers were found in LG14, with a ratio of 33.83% (Table S1).LG12, which harbored 987 markers covering a length of 85.27 cM with the least average interval of 0.09 cM.The max gap on LG12 was only 4.46 cM, which is smaller than that of other linkage groups, while the largest genetic gap was found in LG14, with 23.88 cM genetic distance.The percentage of gap (<5 cM) was 98.40%, indicating that the markers were relatively well-distributed in the map [31].According to a chi-square test (p < 0.05) of the 8171 SNP markers, 429 of which (with a ratio of 5.25%) were segregation distortion markers.The greatest number of segregation distortion markers were found in LG14, with a ratio of 33.83% (Table S1).Haplotype maps were constructed for each individual, and most recombination blocks were identified (Supplementary Figure S1).The average integrity of mapping markers was 99.72%, indicating the accuracy of genotyping and the high quality of the genetic map.Heatmap results indicated a strong linkage relationship between adjacent markers in the linkage group (Supplementary Figure S2).The haplotype map on the LG4 linkage group is shown in Supplementary Figure S3.Haplotype maps were constructed for each individual, and most recombination blocks were identified (Supplementary Figure S1).The average integrity of mapping markers was 99.72%, indicating the accuracy of genotyping and the high quality of the genetic map.Heatmap results indicated a strong linkage relationship between adjacent markers in the linkage group (Supplementary Figure S2).The haplotype map on the LG4 linkage group is shown in Supplementary Figure S3.

QTL Analysis of Morphological Traits
The inclusive composite interval mapping (ICIM) method was used to detect QTLs of morphological traits.A total of 59 associated significant QTLs were identified in two consecutive years, dispersed among 12 LG except for LG3, LG11, and LG14 (Table 4).Each QTL explained the phenotypic variance (PVE) ranging from 6.21% to 17.74%, while the LOD value ranged from 2.75 to 56.83.A total of 33 significant QTLs were detected and had the PVE above 10%.LGs: linkage groups.PVE: the phenotypic variation explained.LOD: the logarithm of odds.The LOD threshold for evaluating the statistical significance (p < 0.05) of each QTL was set by using a 1000 permutations test.The QTLs are labeled with the same letter (a~h) on the upper right and the same background color, indicating the co-localized loci for different traits.
QTLs for spathe length co-localized with spathe width at 169.22 cM on LG1, 55.88 cM on LG10, and 7.34 cM on LG6; whereas QTLs for spadix length, spadix top, and middle diameter were co-localized at the position of 3.44 cM on LG6.The QTL for plant height was co-localized with leaf length at 127.45 cM on LG9.This suggested that some QTLs for morphological traits could be pleiotropy and/or exhibit a tight linkage.

Constructing Linkage Map in Anthurium
Based on RAPD, ISSR, and SRKP molecular markers, one linkage map had been constructed in anthurium before.Due to the limited number of markers and small population size, the genetic map was not saturated enough, but a basic framework was formed [30].In this study, a total of 327,963 SNP markers were identified by applying the SLAFseq technique, of which 131,951 could be successfully encoded and genotyped.After a strict selection, 8171 high-quality SNP markers were used to construct a genetic linkage map, which is the highest number of markers and density to date, with an increase of 7934 markers compared to Venkat et al. [30].
Since 'Pink Champion' and 'Acropolis' were heterozygous clonal cultivars, clonal F 1 progenies contain more alleles at each locus than bi-parental populations derived from two inbred parents, resulting in more significant genetic variation, it is difficult and specific to build a genetic map.By using the software HighMap, which is suitable for crosspollinated (CP) populations [32], the genetic linkage map was constructed successfully with a total length of 2202.27cM distributed in 15 linkage groups, which was consistent with the chromosomes in the Anthurium genus [33].In particular, LG2, LG12, and LG15 were observed with the least inter marker distances, indicating maximum saturation and which might be considered to be recombination hotspots in this population.The mapping rate (mapped marker number/total marker number) was 76.48%.The average interval of 0.30 cM, and gap < 5 cM of 98.40%, indicate that the map was saturated and well-distributed with molecular makers, which were available for QTL analysis.

Segregation Distortion Markers
Segregation distortion is common in constructing a linkage map, where alleles in segregating populations deviate from the expected Mendelian ratio [34].In employing this approach, segregation distortion leads to markers grouping errors and estimating recombination frequency in constructing a linkage map [35,36].However, more studies have reported that only severe segregation distortion could affect mapping; therefore, containing distorted markers could improve the accuracy of marker grouping [37][38][39][40].Therefore, we retained a considerable number of segregation distortion markers for constructing the linkage map in this study.A total of 8171 SNP markers were subjected to the chi-square test (p < 0.05) containing 429 high-quality segregation distortion markers (with a ratio of 5.25%, Table S1), which ensured the accuracy and genomic coverage of the linkage mapping.

QTLs Related to Morphological Traits
The characteristics of spathe, spadix, pedicel, petiole, and leaf are essential components of commercial quality in anthurium.These components are therefore favorable targets for selection in breeding.Generally, the low and compact varieties with small bracts and short pedicels are used as potted flowers, while those with tall and loose plants, larger bracts, and longer pedicels are regarded as cut flowers.Although a few studies have been conducted on the morphology in anthurium [41,42], QTLs associated with morphological traits have not yet been reported.In this study, 15 main morphological traits in F 1 population were measured for consecutive two years, showing a continuous distribution and variability and indicating their quantitative nature of inheritance, which were firstly mapped on the genetic linkage map in anthurium.A total of 59 significant QTLs with individual genetic effects were successfully detected, with each explaining phenotypic variance ranging from 6.21% to 17.74% and the logarithm of odds (LOD) values ranging from 2.75 to 56.83, of which 33 QTLs with PVE > 10% were associated with 13 traits and were designated as major QTLs.However, only one QTL (qPH1) related to plant height and two QTLs (qPtD1, qPtD2) associated with stem diameter with low PVE were identified, indicating that both are complex traits influenced by impacted small effect genes.Notably, most of the QTLs identified in this study are clustered in six locations in the genome (LG1, LG4, LG6, LG8, LG10, and LG12).These six regions contain multiple QTLs controlling different traits.
Despite the huge potential of molecular markers in breeding programs, their implementation in MAS practice has been limited by the lack of information on the stability of QTLs across different environments and within different genetic backgrounds.Here, we present the results from the inheritance analysis of morphological traits within a population derived from a cross between the potted and cut anthurium cultivars over two successive seasons.However, further QTL analyses using multiple populations in larger size and different mapping methods would allow us to precisely position QTL mapping.

Plant Materials and DNA Extraction
The F 1 population consisted of 160 progenies generated from the controlled cross between Anthurium andraeanum Linden cv.'Pink Champion' (♀) and 'Acropolis' (♂).The two cultivars exhibit distinct field performances concerning morphological characters.'Pink Champion' is a dwarf pot flower variety with a smaller blade and spathe, shorter pedicel, and spadix, while 'Acropolis' is relatively giant and widely used as a cut flower variety (Figure 1).Both F 1 individuals and their parents were planted in the ornamental greenhouses of TCGRI-CATAS, Danzhou, China (109 • 42 E/19 • 35 N) and grown in a substrate consisting of 3: 1 (V:V) of composted coco blocks and coarse peat (pH 5.5-6.5)under the conditions of 18-30 • C and 60-80% relative humidity.
For precise phenotyping, the measurements were conducted on the same growth stages for two consecutive years, and all traits were surveyed at least three times.The CV and MPs were analyzed with the software SPSS 26.0.The heatmap and correlation coefficient were created with the software OriginPro 2021 (v9.8.0.200).

SLAF library Construction and Sequencing
Healthy tender leaves (0.5-1.0 g) from the two parents and 160 F 1 individuals were collected separately, and total DNA was isolated with the modified cetyl trimethylammonium bromide (CTAB) method [8].DNA quality was visualized via electrophoresis in 1.0% agarose gels and the concentration was quantified using a DU800 Spectrophotometer (Beckman Coulter, Brea, CA, USA).
According to the estimated size of the anthurium genome (2.8 Gb) and GC content (40.5%),Zea mays were chosen as a reference genome to predict enzymatic digestion, deciding on a combination of endonucleases HinCII and SCal-HF.Subsequently, the digested fragment was subjected to add a 3 end plus a nucleotide and then ligated to a dualindex sequence linker.Polymerase chain reaction (PCR) was performed using diluted restriction-ligation DNA samples and the primer pairs 5 -AATGATACGGCGACCACCGA-3 /5 -CAAGCAGAAGACGGCATACG-3 .PCR products were then purified using the QIAquick Gel Extraction Kit (Qiagen, Hilden, Germany).The target fragments with lengths of 314-414 bp were separated and sequenced on an Illumina HiSeq TM system.To check the accuracy of the library construction and sequencing, Oryza sativa subsp.japonica was selected as a control for the same protocol.To ensure the sequencing quality, the Q30 and GC content were calculated [52].

SNP Markers Detecting and Genotyping
SLAF marker identification and genotyping were performed following previous procedures [24].The clean sequences from all samples were clustered based on sequence similarity.The SLAFs with two to four alleles were defined as potential polymorphic SLAFs, which can develop SNP markers and sort into eight segregation patterns (ab × cd, ef × eg, lm × ll, nn × np, aa × bb, hk × hk, cc × ab, and ab × cc).
To ensure the quality of markers for the linkage map, the rules for SNP marker identification were as follows: (1) SNP markers with a complete degree > 85%.(2) SNP markers with significant segregation distortion (p < 0.05) were removed according to the chi-square test.(3) SLAFs with more than 5 SNP markers were filtered out, as it was considered a high-frequency variant region.(4) SLAFs with a sequencing depth of >25.
(5) Remove redundant markers and the markers (aa × bb) that were not suited to the cross-pollination (CP) population type.

Linkage Map Construction
The construction of a genetic map contains two essential components, grouping and ordering the markers.The MLOD values were calculated between two SNP markers to assign markers on the LGs, and the markers with MLOD < 10 were filtered out before ordering.The HighMap software with the CP option was used for anthurium genetic linkage map construction.The SMOOTH algorithm and the maximum likelihood method were used to correct genotyping errors and order the SNP markers in all LGs.The Kosambi algorithm was used to calculate genetic map distance (cM).Haplotype maps were drawn to detect double crossover populations and genotyping errors.Heatmaps were used to evaluate the relationship of recombination between markers from each LG.

QTL Analysis
The QTL analysis for the average of 15 morphological traits was performed via the ICIM method in GACD V1.0 [53].The logarithm of odds (LOD) threshold for evaluating the statistical significance (p < 0.05) of each QTL was set by using a 1000 permutations test (PT), which is the LOD threshold for evaluating statistical significance.The potential locations of the QTLs were described according to their LOD peak locations and their surrounding regions.If no QTL was detected within the interval for some traits, the LOD score could be manually adjusted low, but it should not be lower than 2.0.A QTL was considered to be a major QTL if it accounted for more than 10% phenotypic variance.

Conclusions
In this study, a F 1 segregation population with 160 individuals derived from the cross Anthurium cv.'Pink Champion' (a potted variety) × 'Acropolis' (a cut variety) was used for constructing a high-density genetic linkage map.Based on SLAF-seq technology, 9134 SNP markers were developed and the first high-resolution genetic linkage map for anthurium was constructed.Fifty-nine QTLs associated with 15 main morphological traits were identified using the ICIM method.The results will lay a foundation for detecting genes related to morphological traits and MAS in anthurium breeding.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/plants12244185/s1, Figure S1.The integrity distribution map of all.Figure S2. the heatmap of LG4. Figure S3.The haplotype map of LG4.Table S1: The markers showing segregation distortion in mapping population.

Figure 2 .
Figure 2. Pearson's correlation coefficient and heatmap of 15 morphological traits in the F1 population.

Figure 3 .
Figure 3.The distribution pattern for 15 traits measured in the mapping population.The subfigure A~O indicate the phenotypic variation of spathe-, spadix-, pedicel-, petiole-, leaf-, and height-related traits among the offspring, respectively.The normal distribution curve in the graph represented the expected percentage with respect to the measurement range of traits.The x-axis indicates the measurement values for each trait; the y-axis indicates frequency.The black and red arrows indicate the mean value of female and male parents, respectively.

Figure 2 .
Figure 2. Pearson's correlation coefficient and heatmap of 15 morphological traits in the F 1 population.

Figure 2 .
Figure 2. Pearson's correlation coefficient and heatmap of 15 morphological traits in the F1 population.

Figure 3 .
Figure 3.The distribution pattern for 15 traits measured in the mapping population.The subfigure A~O indicate the phenotypic variation of spathe-, spadix-, pedicel-, petiole-, leaf-, and height-related traits among the offspring, respectively.The normal distribution curve in the graph represented the expected percentage with respect to the measurement range of traits.The x-axis indicates the measurement values for each trait; the y-axis indicates frequency.The black and red arrows indicate the mean value of female and male parents, respectively.

Figure 3 .
Figure 3.The distribution pattern for 15 traits measured in the mapping population.The subfigure (A-O) indicate the phenotypic variation of spathe-, spadix-, pedicel-, petiole-, leaf-, and height-related traits among the offspring, respectively.The normal distribution curve in the graph represented the expected percentage with respect to the measurement range of traits.The x-axis indicates the measurement values for each trait; the y-axis indicates frequency.The black and red arrows indicate the mean value of female and male parents, respectively.

Figure 4 .
Figure 4.The markers distribution map of eight separation patterns in F 1 progeny.The x-axis stated eight patterns of polymorphic SNPs markers; the y-axis stated the number of markers.

Figure 5 .
Figure 5. Genetic linkage map of the F1 population.The number at the bottom represents each linkage group.The ruler on the left indicates the genetic position in centimorgans (cM).The black bar represents the density of markers (cM/locus).

Figure 5 .
Figure 5. Genetic linkage map of the F 1 population.The number at the bottom represents each linkage group.The ruler on the left indicates the genetic position in centimorgans (cM).The black bar represents the density of markers (cM/locus).

Table 1 .
Descriptive statistics for morphological traits in anthurium parents and F 1 population.
MPs: mid-parent value; Max: maximum value; Min: minimum value; CV: coefficient of variation.Plants 2023, 12, x FOR PEER REVIEW 3 of 15

Table 1 .
Descriptive statistics for morphological traits in anthurium parents and F1 population.

Table 2 .
Statistical sequencing data for the two parents and F 1 progeny.

Table 2 .
Statistical sequencing data for the two parents and F1 progeny.
.Figure 4. The markers distribution map of eight separation patterns in F1 progeny.The x-axis stated eight patterns of polymorphic SNPs markers; the y-axis stated the number of markers.
cM genetic distance containing 127 markers.The most saturated linkage group was

Table 3 .
Marker information for the high-density genetic map.

Table 3 .
Marker information for the high-density genetic map.

Table 4 .
List of major QTLs identified by inclusive composite interval mapping.